metabolite therapy guided by liquid biopsy proteomics ... · h department of pathology & cell...

EBioMedicine 52 (2020) 102636

Contents lists available at ScienceDirect

EBioMedicine

journal homepage: www.elsevier.com/locate/ebiom

Research paper

Metabolite therapy guided by liquid biopsy proteomics delays retinalneurodegeneration

Katherine J. Werta,1, Gabriel Veleza,b,1, Vijaya L. Kanchustambhamc, Vishnu Shankarc,Lucy P. Evansd, Jesse D. Sengilloe, Richard N. Zarec, Alexander G. Bassukd, Stephen H. Tsangf,g,h,Vinit B. Mahajana,i,*a Omics Laboratory, Department of Ophthalmology, Byers Eye Institute, Stanford University, Palo Alto, CA 94304, United StatesbMedical Scientist Training Program, Carver College of Medicine, University of Iowa, Iowa City, IA 52242, United Statesc Department of Chemistry, Stanford University, Stanford, CA 94305, United Statesd Departments of Pediatrics and Neurology, Carver College of Medicine, University of Iowa, Iowa City, IA 52242, United Statese Bascom Palmer Eye Institute, University of Miami Miller School of Medicine, Miami, FL 33136, United Statesf Edward S. Harkness Eye Institute, New York-Presbyterian Hospital, New York, NY 10032, United Statesg Jonas Children’s Vision Care and Bernard & Shirlee Brown Glaucoma Laboratory, Department of Ophthalmology, Vagelos College of Physicians and Surgeons,Columbia University, New York, NY 10032, United StateshDepartment of Pathology & Cell Biology, Stem Cell Initiative (CSCI), Institute of Human Nutrition, Vagelos College of Physicians and Surgeons, Columbia Univer-sity, New York, NY 10032, United Statesi Veterans Affairs Palo Alto Health Care System, Palo Alto, CA 94304, United States

A R T I C L E I N F O

Article History:Received 9 September 2019Revised 26 December 2019Accepted 9 January 2020Available online 3 February 2020

* Corresponding author at: Omics Laboratory, DepartmEye Institute, Stanford University, Palo Alto, CA 94304, U

E-mail address: [email protected] (V.B. Ma1 Katherine J. Wert and Gabriel Velez are co-first autho

https://doi.org/10.1016/j.ebiom.2020.1026362352-3964/Published by Elsevier B.V. This is an open acc

A B S T R A C T

Background: Neurodegenerative diseases are incurable disorders caused by progressive neuronal cell death.Retinitis pigmentosa (RP) is a blinding neurodegenerative disease that results in photoreceptor death andprogresses to the loss of the entire retinal network. We previously found that proteomic analysis of the adja-cent vitreous served as way to indirectly biopsy the retina and identify changes in the retinal proteome.Methods: We analyzed protein expression in liquid vitreous biopsies from autosomal recessive (ar)RPpatients with PDE6A mutations and arRP mice with Pde6ɑ mutations. Proteomic analysis of retina and vitre-ous samples identified molecular pathways affected at the onset of photoreceptor death. Based on affectedmolecular pathways, arRP mice were treated with a ketogenic diet or metabolites involved in fatty-acid syn-thesis, oxidative phosphorylation, and the tricarboxylic acid (TCA) cycle.Findings: Dietary supplementation of a single metabolite, ɑ-ketoglutarate, increased docosahexaeonic acidlevels, provided neuroprotection, and enhanced visual function in arRP mice. A ketogenic diet delayed photo-receptor cell loss, while vitamin B supplementation had a limited effect. Finally, desorption electrospray ioni-zation mass spectrometry imaging (DESI-MSI) on ɑ-ketoglutarate-treated mice revealed restoration ofmetabolites that correlated with our proteomic findings: uridine, dihydrouridine, and thymidine (pyrimidineand purine metabolism), glutamine and glutamate (glutamine/glutamate conversion), and succinic and aco-nitic acid (TCA cycle).Interpretation: This study demonstrates that replenishing TCA cycle metabolites via oral supplementationprolongs retinal function and provides a neuroprotective effect on the photoreceptor cells and inner retinalnetwork.Funding: NIH grants [R01EY026682, R01EY024665, R01EY025225, R01EY024698, R21AG050437,P30EY026877, 5P30EY019007, R01EY018213, F30EYE027986, T32GM007337, 5P30CA013696], NSF grantCHE-1734082.

Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license.(http://creativecommons.org/licenses/by-nc-nd/4.0/)

Keywords:
MetabolomicsMetabolite supplementationTCA cycleKetogenic dietɑ-ketoglutarateDesorption electrospray ionization massspectrometry
ent of Ophthalmology, Byersnited States.hajan).rs.

ess article under the CC BY-NC-ND license. (http://creativecommons.org/licenses/by-nc-nd/4.0/)

1. Introduction

Neurodegenerative diseases are debilitating, chronic disordersassociated with neuronal cell loss in sensory, motor, or cognitive sys-tems [1]. They are typically characterized by abnormal proteostasis,where the cell is unable to maintain regulation of protein folding,

http://crossmark.crossref.org/dialog/?doi=10.1016/j.ebiom.2020.102636&domain=pdf

http://creativecommons.org/licenses/by-nc-nd/4.0/

mailto:[email protected]

https://doi.org/10.1016/j.ebiom.2020.102636

http://creativecommons.org/licenses/by-nc-nd/4.0/


http://www.ScienceDirect.com

http://www.elsevier.com/locate/ebiom

Research in context

Evidence before this study

Retinitis pigmentosa (RP) consists of a group of inherited retinaldegenerative diseases with high genetic heterogeneity, whichcan be caused by mutations in more than 60 genes. Despite theprecise genetic characterization of RP patients, the underlyingmechanisms of photoreceptor cell death is not well understood.Our previous studies have shown that changes in the retinalproteome during disease can be detected through proteomicanalysis of vitreous biopsies, as proteins from molecular path-ways affected during disease pathogenesis are elevated in thevitreous fluid.

Added value of this study

We applied a liquid biopsy proteomics approach to identify theproteins lost from the retina and elevated in the vitreous duringRP progression. We then used these results to determine themolecular pathways most affected at the onset of disease. Wefound that metabolic and cellular energy pathways were highlydisrupted at disease onset. Next, we replenished metabolitesfound within these disrupted cellular pathways and determinedthat some metabolites, specifically those involved in the tricar-boxylic acid (TCA) cycle, can provide neuroprotection to the pho-toreceptor cells in our RP mouse model.

Implications of all the available evidence

This study provided key insight into the cellular pathwaysinvolved in retinal degeneration. We found that cellular metabo-lism and energy pathways are highly disrupted at the onset ofRP, and that the delivery of metabolites within these critical path-ways in the arRP preclinical mouse model showed efficaciousneuroprotective effects on the inner retina. The results from thisstudy lay the groundwork for future experiments to address howthe TCA cycle and aerobic metabolism influence the photorecep-tors and their signaling to the inner retinal network, and thusimprove upon neuroprotective approaches targeting photorecep-tor cell metabolism. This work also suggests that directly target-ing the TCA cycle may be a beneficial approach for future RPtherapy.

2 K.J. Wert et al. / EBioMedicine 52 (2020) 102636

translation, and degradation pathways. These defects in proteostasislead to processes of cell death such as apoptosis and disrupted signal-ing mechanisms within the cell from oxidative stress, glutamate-mediated cytotoxicity, neuroinflammation, mitochondrial dysfunc-tion, and increased endoplasmic reticulum (ER) stress [2�7]. How-ever, the early factors involved in neuronal cell death are still notunderstood. Retinitis pigmentosa (RP) consists of a group of inheritedretinal neurodegenerative diseases with high genetic heterogeneity,which can be caused by mutations in more than 60 genes. RP affectsapproximately 1.5 million people worldwide, and approximately 10%of Americans carry a recessive mutant RP allele. Despite precisegenetic characterization of RP patients, identification of causativemutations has had limited impact on therapy. To date, the only treat-ment for RP is supplementation with vitamin A (15,000 IU/day), butits clinical value is controversial, it does not cure disease, and vitaminA can have adverse effects for other retinal neurodegenerations suchas Stargardt macular dystrophy [8]. There have been recent advancesto treat RP patients using prosthetics or gene therapy approaches[9,10], but only end-stage RP or patients with the RPE65 gene muta-tion, respectively, are currently eligible for such therapies. Thus, thereis a critical need to understand the underlying mechanisms of

photoreceptor death and provide a general therapy independent ofspecific gene mutations. Delaying photoreceptor cell degeneration,even without a complete cure, can significantly increase the qualityof life for RP patients and prolong their ability to live independently.

There is accumulating evidence linking neurodegeneration toenergy metabolism, particularly in age-related neurological disorderslike Huntington’s, Alzheimer’s, and Parkinson’s disease [1]. Neuronalaging is often accompanied by metabolic and neurophysiologicchanges that are associated with impaired function. In the eye, retinalphotoreceptors are metabolically robust and require high rates ofaerobic ATP synthesis. Defects in retinal cellular metabolism are asso-ciated with age-related degenerative retinal diseases, such as roddystrophy and RP. We have previously shown, using proteomics, thatthe representation of metabolic pathways and enzymes in the humanretina varies by anatomic location (i.e. the peripheral, juxta-macular,and foveomacular regions) [11]. In addition to anatomical differencesin metabolic pathway representation, energy consumption in the ret-inal cell layers is highly compartmentalized. It has been demon-strated that the mammalian inner retina metabolizes a majority(»70%) of its glucose via the tricarboxylic acid (TCA) cycle and oxida-tive phosphorylation (OXPHOS). In contrast, the outer retina metabo-lizes »60% of its glucose through glycolysis [12,13]. Thus, defects inaerobic metabolism could preferentially affect inner retinal function.In support of this, photoreceptor cells of the retina are highly suscep-tible to glycolytic inhibition (via sodium iodoacetate injection) andmature ‘RP’ rats lacking a majority of their photoreceptors display an»50% reduction in glycolytic activity compared to normal rat retinas[14,15]. Glucose metabolism plays other non-energetic roles in theretina in addition to meeting the high energy demands of photore-ceptors [16,17]. For example, glucose metabolism by the pentosephosphate pathway leads to the production of intracellular glutathi-one, which prevents neuronal cell death through redox inactivationof cytochrome c-mediated apoptosis pathways [18]. Glucose metabo-lism also feeds into fatty acid synthesis and lipid metabolism path-ways, which are responsible for maintaining proper membrane lipidcomposition in neuronal cells [19]. Disruption of lipogenesis (bydeleting fatty acid synthase) in the neural retina leads to abnormalsynaptic structure and apoptosis, resulting in a phenotype resem-bling RP [20]. Thus, loss of these key metabolic pathways andenzymes would result in abnormal nutrient and metabolite availabil-ity, and impaired visual function.

It has been shown that the risk of neuronal cell death is constant,regardless of the underlying cause [21]. Therefore, pathways mustexist that can be targeted to generally reduce or protect against neu-ronal cell death. A potential therapeutic approach for RP is to repro-gram metabolic pathways that are altered during rod dystrophy andRP progression. In one such approach, we found that knocking downsirtuin-6 (SIRT6) to reprogram retinal cells toward anabolism delayedphotoreceptor cell degeneration in a mouse model of RP [22]. For thisstudy, we applied a liquid biopsy proteomics approach to identify themolecular pathway targets affected during RP progression. Using pro-teomics, we have previously shown that molecular changes in theneural retina can be detected through proteomic analysis of humanand mouse vitreous biopsies, since the retina deposits proteins intothe adjacent vitreous extracellular matrix in retinal disease [23, 24].In this study, we collected vitreous biopsy samples from patientswith autosomal recessive retinitis pigmentosa (arRP) carrying muta-tions in PDE6A, a gene encoding the catalytic ɑ-subunit of the rod-specific cGMP-dependent phosphodiesterase (PDE6), and from micecarrying a Pde6ɑ mutation. Based on proteomic analyses, we targeteddiseased metabolic pathways to investigate treatments for RPpatients and found key metabolites that prolong photoreceptor cellsurvival and visual function. We further compared the distribution ofthese metabolites using desorption electrospray ionization massspectrometry imaging (DESI-MSI) between untreated Pde6ɑmice andmetabolite-treated mice.

K.J. Wert et al. / EBioMedicine 52 (2020) 102636 3

2. Materials and methods

2.1. Study approval

The study protocol was approved by the Institutional Review Boardfor Human Subjects Research (IRB) at Columbia University and Stan-ford University, was HIPAA compliant, and adhered to the tenets of theDeclaration of Helsinki. All subjects underwent informed consent forstudy participation. All experiments were performed in accordancewith the ARVO Statement for the Use of Animals in Ophthalmic andVisual Research and were all approved by the Animal Care and UseCommittee at Columbia University and Stanford University.

2.2. Human patient imaging

Clinical examination and testing were performed as previouslydescribed [25]. Autofluorescence (AF) images were obtained using aTopcon TRC 50DX camera (Topcon, Pyramus, NJ, USA). Optical coher-ence tomography imaging was obtained from the spectral-domainHeidelberg HRA2 Spectralis, version 1.6.1 (Heidelberg Engineering,Inc, Vista, CA, USA). Genetic testing was performed as previouslydescribed [26].

2.3. Human vitreous sample collection

Pars plana vitrectomy for vitreous biopsy was performed as previ-ously described [27]. We used a single-step transconjunctival 23-gauge trocar cannula system (Alcon Laboratories Inc, Fort Worth, TX).A light pipe and vitreous cutter were inserted into the mid vitreousand the cutter was activated for 30 s without infusion. An undiluted1.0-cc sample of vitreous was then manually aspirated into a 3-ccsyringe. Vitreous samples were immediately centrifuged in the oper-ating room at 15,000 £ g for 5 min at room temperature to removeparticulate matter, and samples were then stored at �80 °C.

2.4. Mouse lines and husbandry

C57BL/6J-Pde6anmf363/nmf363 (RRID: MGI:3828994), with a D670Gmutation, herein referred to as Pde6aD670G mice, were obtained fromthe Jackson Laboratory (Bar Harbor, ME, USA). Pde6aD670G are co-iso-genic in the C57BL/6J (B6) background; therefore, age matched B6mice (RRID: IMSR_JAX:000664) were used as experimental controls.Mice were bred and maintained at the facilities of Columbia Universityand Stanford University. Animals were kept on a light�dark cycle(12 h�12 h). Food and water were available ad libitum throughout theexperiment, whether treated or untreated. The ketogenic diet was pro-vided by BioServ (Ketogenic Diet, AIN-76A-Modified, High-Fat, Paste;Flemington, NJ) in place of standard chow. The ketogenic diet wasstored refrigerated and changed every other day for the mice duringthe duration of the experiment. Vitamins B2 (riboflavin, Sigma Aldrich)and B3 (nicotinamide, Sigma Aldrich) were provided at a concentrationof 2.5 g/L in the mouse drinking water. ɑ-KG (alpha-ketoglutaric acid,Sigma Aldrich) was provided at a concentration of 10 g/L in the mousedrinking water. Melatonin (Sigma Aldrich) was provided at a concen-tration of 80 mg/L in the mouse drinking water. Resveratrol (SigmaAldrich) was provided at a concentration of 120 mg/L in the mousedrinking water. All water bottles were covered in foil to reduce lightexposure and changed weekly during the duration of the experimentfollowing approved institutional guidelines.

2.5. Mouse autofluorescence/infrared imaging

Autofluorescence/infrared (AF/IR) fundus imaging was obtainedwith the Spectralis scanning laser confocal ophthalmoscope (OCT-SLO Spectralis 2; Heidelberg Engineering, Heidelberg, Germany) aspreviously described [28,29]. Pupils were dilated using topical 2.5%

phenylephrine hydrochloride and 1% tropicamide (Akorn, Inc., Lake-forest, IL, USA). Mice were anesthetized by intraperitoneal injectionof 0.1 mL /10 g body weight of anesthesia [1 mL ketamine�100 mg/mL (Ketaset III, Fort Dodge, IA, USA) and 0.1 mL xylazine—20 mg/ mL(Lloyd Laboratories, Shenandoah, IA, USA) in 8.9 mL PBS]. AF imagingwas obtained at 488-nm absorption and 495-nm emission using a55° lens. IR imaging was obtained at 790 nm absorption and 830 nmemission using a 55° lens. Images were taken of the central retina,with the optic nerve located in the center of the image.

2.6. Electroretinography (ERG)

Mice were dark-adapted overnight, manipulations were con-ducted under dim red-light illumination, and recordings were madeusing Espion or Celeris ERG equipment (Diagnosys LLC, Littleton, MA,USA) as previously described [28,29]. Briefly, adult B6 control micewere tested at the beginning of each session to ensure equal readoutsfrom the electrodes for both eyes before testing the experimentalmice. Pupils were dilated using topical 2.5% phenylephrine hydro-chloride and 1% tropicamide (Akorn Inc., Lakeforest, IL, USA). Micewere anesthetized by intraperitoneal injection of 0.1 mL/10 g bodyweight of anesthesia [1 mL ketamine 100 mg/mL (Ketaset III, FortDodge, IA, USA) and 0.1 mL xylazine 20 mg/mL (Lloyd Laboratories,Shenandoah, IA, USA) in 8.9 mL PBS]. Body temperature was main-tained at 37 °C during the procedure. Both eyes were recorded simul-taneously, and responses were averaged for each trial. Responseswere taken from the Espion/Celeris readout in microvolts (mV). Sta-tistical significance was set at p < 0.05, and 1-way ANOVA followedby Tukey’s multiple comparison’s test was used to compare alltreated, untreated, and wild-type groups.

2.7. Outer nuclear layer quantification

Mice were sacrificed, and the eyes enucleated as previouslydescribed [30,31]. Eyes were embedded in paraffin, sectioned, andstained with hematoxylin and eosin by Excalibur Pathology, Inc.(Norman, OK), before being visualized by light microscopy (Leica DM5000B, Leica Microsystems, Germany). Quantification of photorecep-tor nuclei was conducted on several sections that contained the opticnerve, as follows: the distance between the optic nerve and the cili-ary body was divided into three, approximately equal, quadrants oneach side of the retina. The number of nuclei in four columns werecounted within each single quadrant. These counts were then used todetermine the average thickness of the ONL for each individual ani-mal and within each quadrant of the retina, spanning from the ciliarybody to the optic nerve head (ONH) and out to the ciliary body. Sec-tioning proceeded along the long axis of the segment, so that eachsection contained upper and lower retina as well as the posteriorpole. Statistical analysis was performed on the average ONL thicknessof each individual animal, taking into account all cell counts for allquadrants, and Student’s t-test was performed with statistical signifi-cance set at p < 0.05.

2.8. Immunocytochemistry

Unfixed enucleated eyes were embedded in optimal cutting tem-perature (OCT) medium and frozen using dry ice. Sagittal sections(10 mm thick) of the ipsilateral retina were collected onto superfrostplus glass slides. OCT-embedded sections were placed in precooled4% paraformaldehyde and brought to room temperature for 20 minto fix the tissue. Sections were blocked for 1 hour with 3% goatserum/0.1% Triton X-100 (cat. no. T8532; Sigma-Aldrich, St. Louis,MO) in PBS. Primary antibody against glial fibrillary acidic protein(GFAP; 1:400; cat. no. Mab360; Millipore, Burlington, MA; RRID:AB_11212597) was diluted in 3% goat serum and PBS overnight at4 °C. The slides were washed with PBS (3 £ 5 min) and incubated


with the secondary antibody GAM647 (1:300; cat. no. A-32728;Thermo Fisher Scientific; RRID: AB_2633277) in 3% goat serum in PBSfor 1 hour at room temperature. Samples were washed with PBS(3 £ 5 min), incubated with DAPI (4 mg/mL; cat. no. D9564; Sigma-Aldrich) for 10 min at room temperature and washed with PBS(3 £ 5 min). Slides were mounted with ProLong Gold AntifadeMountant (cat. no. P36934; Thermo Fisher Scientific; RRID:SCR_015961) followed by imaging with confocal microscopy at 25£(SP8 confocal, Leica Microsystems, Wetzlar, Germany).

2.9. Mouse vitreous and retina sample collection

The vitreous and retina from 12 mouse eyes were eviscerated asdescribed previously [32]. Briefly, scleral tissue posterior to the lim-bus was grasped with 0.22 forceps and a microsurgical blade wasused to make a linear incision in the cornea from limbus to limbus. Afine curved needle holder was inserted behind the lens toward theposterior aspect of the globe. The needle holder was partially closedand pulled forward pushing the lens through the corneal incisionwhile leaving the eye wall intact. The vitreous was partially adherentto the lens. The lens-vitreous tissue was then placed into a filteredcentrifugation tube containing 20 mL of protease inhibitor cocktail(Roche) dissolved in PBS. The fine curved needle holder was placedas far posterior to the globe as possible, near the optic nerve. The nee-dle holder was partially closed, and pulled forward, pushing the ret-ina forward through the corneal incision. The vitreous appeared as atranslucent gel adherent to the retina. The retina-vitreous tissue wasplaced into the filtered centrifuge tube containing the lens-vitreoustissue. The filtered centrifuge tube was spun at 14,000 £ g for 12 minand the eluent (vitreous) was collected. Retina was separated fromthe lens tissue remaining in the filtered centrifuge tube and collectedin PBS containing protease inhibitor cocktail.

2.10. Protein extraction and digestion

The received vitreous and retina samples were diluted in 2% SDS,100 mM Tris-HCl (pH 7.6), 100 mM DTT to approximately 0.5 mL vol-ume and heated at 95 °C for 10 min. Each sample was then briefly vor-texed and sonicated for 10 s using a probe-tip sonicator (OmniInternational). The samples were then returned to incubate at 95 °Cfor an additional 10 min. Samples were then transferred to a 30 K Ami-con MWCO device (Millipore) and centrifuged at 16,100 £ g for30min. Then 400mL of 8 M urea, 100 mM Tris-HCl (pH 7.6) was addedto each device and centrifuged as before and the filtrate discarded. Thisstep was repeated once. Then 400 mL of 8 M urea, 100 mM Tris-HCl(pH 7.6), 15 mM iodoacetamide was added to each device and incu-bated in the dark for 30 min. The samples were then centrifuged asbefore, and the filtrate discarded. Then 400 mL of 8 M urea, 100 mMTris�HCl (pH 7.6) was added to each device and centrifuged as beforeand the filtrate discarded. This step was repeated once more. Then400 mL of 2 M urea, 100 mM Tris-HCl (pH 7.6) was added to eachdevice along with 2.5 mg trypsin. The devices incubated overnight ona heat block at 37 °C. The devices were then centrifuged, and the fil-trate collected. Then 400 mL 0.5 M NaCl was added to each device andcentrifuged as before. The filtrate was added to the previously col-lected filtrate.

2.11. Peptide desalting and fractionation

Digested peptides were desalted using C18 stop-and-go extrac-tion (STAGE) tips. Briefly, for each sample a C18 STAGE tip was acti-vated with methanol, then conditioned with 75% acetonitrile, 0.5%acetic acid followed by 0.5% acetic acid. Samples were loaded ontothe tips and desalted with 0.5% acetic acid. Peptides were eluted with75% acetonitrile, 0.5% acetic acid and lyophilized in a SpeedVac(Thermo Savant) to dryness, approximately 2 hours. Peptides were

fractionated using strong anion exchange (SAX) STAGE tips. Briefly,for each sample a SAX STAGE tip was activated with methanol, thenconditioned with Britton�Robinson buffer (BRB), pH 3.0 followed byBRB (pH 11.5). Peptides were loaded onto the tips and the flow-through collected followed by and five additional fractions by subse-quent application of BRB at pH 8.0, 6.0, 5.0, 4.0 and 3.0. Each fractionwas desalted using a C18 STAGE tip and lyophilized as describedabove.

2.12. Liquid chromatography-tandem mass spectrometry (LC-MS/MS)

Each SAX fraction was analyzed by LC-MS/MS. LC was performedon an Agilent 1100 Nano-flow system. Mobile phase A was 94.5%MilliQ water, 5% acetonitrile, 0.5% acetic acid. Mobile phase B was80% acetonitrile, 19.5% MilliQ water, 0.5% acetic acid. The 150 min LCgradient ran from 5% A to 35% B over 105 min, with the remainingtime used for sample loading and column regeneration. Sampleswere loaded to a 2 cm £ 100 mm I.D. trap-column positioned on anactuated valve (Rheodyne). The column was 13 cm £ 100 mm I.D.fused silica with a pulled tip emitter. Both trap and analytical col-umns were packed with 3.5 mm C18 (Zorbax SB, Agilent). The LC wasinterfaced to a dual pressure linear ion trap mass spectrometer (LTQVelos, Thermo Fisher) via nano-electrospray ionization. An electro-spray voltage of 1.5 kV was applied to a pre-column tee. The massspectrometer was programmed to acquire, by data-dependent acqui-sition, tandem mass spectra from the top 15 ions in the full scan from400�1,400m/z. Dynamic exclusion was set to 30 seconds.

2.13. LC-MS/MS data processing and library searching

Mass spectrometer MS data files were converted to mzXML formatand then to mgf format using msconvert in ProteoWizard andmzXML_to_mgf v4.4, rev 1, respectively. Peak list data were searchedusing two algorithms: OMSSA v2.1.9 and X!Tandem TPP v4.4, rev 1.The UniProt mouse or human protein sequence library was used in atarget-decoy format (derived from UniProt as of 10/22/2012). The mgffiles were searched using OMSSA with precursor settings of +/� 2.0 Daand fragment tolerance of +/� 0.8 Da. In X!Tandem, precursor settingswere set to �1 Da and +3 Da and the fragment tolerance to 0.8 Da. Thenumber of missed cleavages was set to 1; the fixed modifications wereCarbamidomethyl (C); the variable modifications were Oxidation (M).XML output files were parsed using TPP v4.4, rev 1 in the programsPeptideProphet, iProphet, and ProteinProphet. Proteins were requiredto have 2 or more unique peptides with E-value scores of 0.01 or less.Relative quantitation was performed by spectral counting. Data werenormalized based on total spectral counts (hits) per sample.

2.14. LC-MS/MS data analysis and visualization

Results (i.e. protein spectral counts) were saved in Excel as .txtformat and were uploaded into the Partek Genomics Suite 6.5 soft-ware package as described previously [23,24]. The data was normal-ized to log base 2 and compared using 1-way ANOVA analysis. Allproteins with non-significant (p > 0.05) changes were eliminatedfrom the table. The significant values were mapped using the ‘clusterbased on significant genes’ visualization function with the standardi-zation option chosen. Results of the 1-way ANOVA analysis were dis-played as a heatmap with relative protein expression between theexperimental groups represented on a log scale. Each column of theheatmap represents an individual sample (n = 3 for each group) andeach row represents an individual protein. Orange indicates highexpression, while green/black indicates low or no expression. PAN-THER Pathway Analysis was utilized to determine the most signifi-cant molecular pathways affected by the proteins present in eachgroup [33]. Gene ontology (GO) analysis was also performed in PAN-THER. Pie charts were created for the visualization of GO distributions


within the list of proteins for each GO term category including biolog-ical process, molecular function, and cellular component.

2.15. Desorption electrospray ionization mass spectrometry imaging(DESI-MSI)

Mouse retinas were dissected and frozen in optimal cutting tem-perature (OCT) media. Cryosections (10-mm) were collected (CryostatMicrosystem, Leica Biosystems) with at least three retinal sectionsper SuperFrost Plus Slide (Fisher Scientific), with at least three slidescollected per mouse (at least 9 retinal sections per mouse). Briefly,DESI-MSI was performed in the negative ion mode (�5 kV) from m/z50�1,000, using LTQ-Orbitrap XL mass spectrometer (Thermo Scien-tific) coupled to a home-built DESI-source and a two-dimensional(2D) motorized stage. The retinal tissues were raster scanned underimpinging charged droplets using a 2D moving stage in horizontalrows separated by a 200-mm (spatial resolution) vertical step. Theimpinging primary charged droplets were generated from the elec-trospray nebulization of a histologically compatible solvent system,1:1 (vol/vol) dimethylformamide/acetonitrile (DMF/ACN, flow rate1 mL/min). The electrospray nebulization was assisted by sheath gasnitrogen (N2, 170 psi) and a high electric field of �5 kV. The spatialresolution of DESI-MSI is defined by the spray spot size of »200 mm.DESI-MSI of all tissue samples were carried out under identicalexperimental conditions, such as spray tip-to-surface distance»2 mm, spray incident angle of 55°, and spray-to-inlet distance»5 mm. MSI data acquisition was performed using XCalibur 2.2 soft-ware (Thermo Fisher Scientific Inc.). XCalibur 2.2 mass spectra files(.raw files) were converted to .mzML format files using msconvert;then imzMLConverter (version 1.3.0) was used to combine .mzMLfiles into .imzML format files, readable by another software MSiR-eader (version v1.00). The 2D chemical maps of molecular ions wereplotted using Biomap. To identify significantly differentiallyexpressed metabolites, XCalibur raw data files were converted to csvfiles for statistical analysis. Raw csv data was imported to the R lan-guage for further processing. To account for random effects influenc-ing mass spectrometry data, the intensity of each metabolite wasnormalized by the total ion current for the corresponding pixel. Next,a nearest neighbor clustering method was used to collect the pixelintensities corresponding to the nearest molecular ion peak. To deter-mine which metabolites could distinguish Pde6aD670G mice fromɑ-KG treated mice, we applied the SAM (Significance Analysis ofMicroarrays) technique using the samr package in R [34]. The metab-olites were filtered according to a false discovery rate (FDR) cutoff of5%. SAM identified 913 potentially significant metabolites(FDR < 0.05), of which 377 metabolites had an absolute fold changegreater than 1. Although SAM returned 377 metabolites with notablefold change, to help interpret the biological role of metabolites, werestricted our analysis to peaks for which tandem-MS and high massresolution analyses were performed by using the LTQ-Orbitrap XL(Thermo Scientific). The relative intensity is equal to the intensity ofthe metabolite normalized by the total ion current (sum of all inten-sity values from the detected ions).

2.16. Metabolite/lipid identification

For lipid and metabolite identification, tandem-MS and high-massresolution analyses were performed using the LTQ-Orbitrap XL(Thermo Scientific). The retinal tissue sections were dissolved in500 mL of methanol: water (80:20) for metabolite extraction. Theundissolved tissue was centrifuged down by 4000 £ g for 5 min. Theresultant clear solution was sprayed into mass spectrometer for tan-dem-MS and high-mass resolution analyses, at a flow rate of 2 mL/min, sheath gas pressure (N2, 120 psi), and the spray voltage �5 kV.For the MS/MS studies, a normalized collision energy of 20�40%was applied. The metabolites and lipids were assigned based on

high-mass resolution analysis, isotope distribution, and tandem-MSfragmentation patterns. In the case when several detectable fragmen-tation patterns were generated by isomeric parent ions, the mostintense fragmentation peaks were used to assign the molecule iden-tity listed in SI Appendix. The LipidMaps (www.lipidmaps.org/),MassBank (www.massbank.jp), Metlin (https://metlin.scripps.edu/),and the human metabolome database (HMDB) were used for themetabolite identification. The detected species were mostly deproto-nated small metabolites related to the tricarboxylic acid (TCA) cycle,and deprotonated lipids including free fatty acids (FAs), fatty aciddimers, phosphatidic acids, and glycerophospholipids.

2.17. Statistical analysis

Differences in experimental groups were determined by the Stu-dent’s t-test as appropriate, or by one-way ANOVA followed byTukey’s post-hoc multiple comparison’s test (see individual methodssection). A p value < 0.05 was considered significant. Measurementswere done blinded to experimental group. All error bars representstandard deviation unless otherwise listed.

3. Results

3.1. Proteomics of the human vitreous highlight changes occurring inthe neural retina during RP disease progression

The proband presented to us at age 38with end-stage arRP. Genetictesting revealed p.R102C/p.S303C mutations in the PDE6A gene (II:5;Fig. 1(a)). Mutations in PDE6 contribute to a significant fraction of RPcases (7�9%; OMIM: 180071) [35]. In comparison to a normal control(Fig. 1(b)�(d)), fundus autofluorescence (AF) revealed high-densitycentral AF rings (Fig. 1(e)�(f)) and optical coherence tomography(OCT) revealed thinning of the neuronal cell layers (Fig. 1(g)). Addition-ally, the proband’s affected brother (II:4; Fig. 1(a)) displayed a similardisease stage with hyper-autofluorescent rings on AF imaging (Fig. 1(h)�(i)). OCT imaging similarly revealed thinning of retinal cell layersand cystoid macular edema (Fig. 1(j)). The affected brothers bothdeveloped significant epiretinal membranes (ERMs) and underwentvitrectomy surgery. Liquid vitreous biopsies were collected at the timeof surgery (Fig. 2(a) -(b), Table S1). ERM vitreous samples from twopatients without RP were used as comparative controls since the twoarRP patients underwent vitrectomy surgery for ERM removal (Fig. 2(a)-(b)). Vitreous samples were analyzed using shotgun liquid chroma-tography-tandem mass spectrometry (LC-MS/MS) to determine prote-omic content (Table S2).

It is rare to obtain arRP vitreous samples as the indication for vit-rectomy surgery in these patients is exceptional. Despite the limitationof small sample size, descriptive semi-quantitative analysis can pro-vide insight into underlying human disease mechanisms in a rarepatient population, and there are no human arRP proteomic reports todate. We therefore performed label-free relative quantification of pro-teins identified in arRP vitreous relative to ERM, which was defined asthe reference group. We detected 612 proteins in arRP vitreous by LC-MS/MS (Table S3). We compared protein expression between arRPand ERM vitreous: there were 110 elevated proteins (with a fold-change �2) in arRP vitreous compared to ERM controls (Fig. 2(c)). Weobserved a significant fraction (36%) of elevated, intracellular proteinsin arRP vitreous despite the extracellular location and acellular compo-sition of the vitreous. We previously reported that a significant fractionof proteins in the normal human vitreous are intracellular, particularlyin the vitreous core, suggesting that surrounding tissues release pro-teins into the adjacent vitreous [36]. Among these elevated intracellu-lar proteins were fatty acid synthase (FASN) and insulin-like growthfactor-binding protein 2 (IGFBP2). FASN is involved in the synthesis oflong-chain fatty acids from acetyl-CoA and NADPH. IGFBP2 hasbeen shown to promote fibroblast differentiation and angiogenesis

http://www.lipidmaps.org/

http://www.massbank.jp

https://metlin.scripps.edu/

Fig. 1. Two patients with autosomal recessive retinitis pigmentosa carrying muta-tions in PDE6A: (a) Pedigree of a family with autosomal recessive retinitis pigmentosa(arRP) caused by mutations in the PDE6A gene. The proband is denoted by the arrow.(b�c) Fundus autofluorescence (488 nm) of a control human eye. (d) Spatially corre-sponding SD-OCT scans through the fovea of a control human eye. (e�f) Fundusautofluorescence (488 nm) of the proband (II:5) revealed confluent areas of hypo-autofluorescence, suggesting RPE loss throughout the periphery, and a central ring ofhyper-autofluorescence overlying the parafovea. (g) Spatially corresponding SD-OCTscans through the fovea confirmed marked peripheral outer retinal thinning and RPEloss. Despite macular edema, the fovea was relatively preserved, with a distinguish-able ellipsoid zone. (h�i) Fundus autofluorescence of the affected brother (II:4)revealed similar confluent areas of hypo-autofluorescence and a central ring ofhyper-autofluorescence overlying the parafovea. (j) Spatially corresponding SD-OCTscans showed similar marked peripheral outer retinal thinning and RPE loss.


(Fig. 2(c)) [37,38]. There were 135 depleted proteins (with a fold-change � �2) in arRP vitreous compared to ERM controls. Amongthese proteins were microfibrillar-associated protein 4 (MFAP4) andamyloid beta precursor like 2 (APLP2; Fig. 2(c)). MFAP4 plays a role inmaintaining integrin activation, which is important for maintainingRPE-cell adhesion to Bruch’s membrane [39]. Interestingly, APLP2 lossor dysfunction has been shown to result in abnormal amacrine andhorizontal cell differentiation [39]. These results suggested that the

observed proteomic changes in human arRP vitreous could reflectmolecular changes in the neural retina. However, the human patientproteomes reflected samples collected at a very late stage of disease,long after significant photoreceptor death andmay not reflect the criti-cal cellular pathways altered at the onset of neurodegeneration. Wetherefore chose to gain further insight into the earlier disease stagesusing a preclinical model of arRP.

3.2. Proteomic analysis of Pde6ɑD670G retina and vitreous identifiesalterations in global protein expression that may precede cell loss andthe RP clinical phenotype

We utilized the Pde6ɑD670G mutant mouse to determine whetherany of the cellular pathways found in our human patient proteomeswere disrupted at the onset of rod dystrophy, before the clinical RPphenotype is detectable. The Pde6ɑD670G phenotype closely resemblesthat of patients with arRP. These mice develop early onset severe reti-nal degeneration, characterized by photoreceptor death, retinal vesselattenuation, pigmented patches, ERG abnormalities, and white retinalspotting [29]. The mice display a loss of rod photoreceptor visual func-tion by ERG analysis as early as one month of age and full loss of globalvisual function on ERG analysis by twomonths of age (Fig. 3(a)). In par-ticular, Pde6ɑD670G mice do not display histologically detectable photo-receptor degeneration until after postnatal day 14 (P14; Fig. 3(b)), soretina and vitreous samples were collected from these mice at diseaseonset (P15), mid-stage (P30) and late-stage disease (P90; Table S4).Retina and vitreous samples for wild-type and Pde6ɑD670G mice weretrypsinized and underwent LC-MS/MS analysis (Table S5-6). Notablythere were fewer unique proteins in the Pde6ɑD670G mouse retina at90 days (484 § 63 individual proteins) than 15 days (1,135 § 55 indi-vidual proteins), suggesting a decrease in protein expression as neuro-degeneration progresses and/or increased degradation of theseproteins (Table S7). Conversely, we noticed there were more uniqueproteins in the Pde6ɑD670G mouse vitreous at 15 days (364 § 30 indi-vidual proteins) compared to wild-type vitreous (202 § 9 individualproteins), suggesting an increase in vitreous protein expression atearly stages of neuronal cell death (Table S8).

Retina and vitreous protein levels (with two or more spectra) fromPde6ɑD670G and wild-type mice were compared using 1-way ANOVAand heatmap clustering (Table S9). A total of 1,384 proteins were differ-entially expressed in the Pde6ɑD670G retina and vitreous samples com-pared to controls (p < 0.05 [1-way ANOVA]; Fig. 3(c)). There were 303proteins highly-expressed in Pde6ɑD670G vitreous at day 15 that werenot present in the control retina or vitreous (Fig. 3(c)). A majority ofthese proteins were intracellular and represented protein synthesis,gene expression, DNA replication, recombination, and repair pathways.Of these 303 proteins, there were 16 downstream effectors of the CD3receptor: Coronin-1A (CORO1A), DEAD box protein 19A (DDX19A),eukaryotic translation initiation factor 4E (EIF4E), GABA receptor-associ-ated protein-like 2 (GABARAP12), hemopexin (HPX), E3 ubiquitin-pro-tein ligase Huwe1 (HUWE1), Lon protease homolog (LONP1), acyl-protein thioesterase 1 (LYPLA1), metastasis-associated protein (MTA2),alpha-N-acetylgalactosaminidase (NAGA), purine nucleoside phosphor-ylase (PNP), RNA polymerase II subunit B1 (POLR2A), ubiquitin receptor(RAD23B), thiosulfate sulfurtransferase (TSTA2), thioredoxin reductase1 (TXNRD1), and alanine-tRNA ligase (AARS). The CD3 co-receptor isexpressed on T-cells and activates cytotoxic and helper T-cells [40]. Thepresence of downstream CD3 effectors in the Pde6ɑD670G vitreous sug-gests that lymphocytes may be active at early stages of the disease. Fur-ther investigation of the RP immune response is required to validatethis observation. There were 360 proteins in the control retina thatwere absent in the degenerating (Pde6ɑD670G) retina at P15. These 360proteins were absent in the control vitreous but, surprisingly, weredetected in the diseased (Pde6ɑD670G) vitreous (Fig. 3(c)). Together, thisindicates that these proteins have changed their location from the retinato the vitreous in the arRP mice as compared to wild-type controls.

Fig. 2. Proteomic analysis of PDE6A human vitreous. (a�b) Pars plana vitrectomy and epiretinal membrane (ERM) removal procedure for an PDE6A (arRP) patient (II:5 in Fig. 1). Avitreous biopsy was collected from this patient and the affected brother. Protein content was analyzed by LC-MS/MS. The vitreous cutter is denoted by the letter ‘v.’ (c) Protein spec-tral counts were analyzed by 1-way ANOVA. Results are represented as a volcano plot. The horizontal axis (x-axis) displays the log2 fold-change value (arRP vs. ERM) and the verticalaxis (y-axis) displays the noise-adjusted signal as the -log10 (p-value) from the 1-way ANOVA analysis. Increased levels of FASN and IGFBP2 are denoted by the orange circles. Bluecircles indicate decreased levels of MFAP4 and APLP2 in arRP vitreous compared to controls.


To obtain a global view of the Pde6ɑD670G retina and vitreous pro-teomes at different stages of degeneration, a gene ontology (GO)analysis of highly represented proteins in each group was performed(statistically-significantly elevated proteins in each group; p < 0.05[1-way ANOVA]). The highest represented categories in the retinawere cellular process, catalytic activity, and intracellular regions(Fig. S1). Interestingly, there was a decrease in the number of proteinswith transporter activity in the Pde6ɑD670G retina at early stage neu-ronal cell death, consistent with a loss of protein expression at thisstage (Fig. S1). When comparing the four total retinal protein profiles,the GO summaries for the control retina were similar to thePde6ɑD670G retina at P30 and P90. This result suggests that the com-parison of the overall ontological distribution of proteins (withrespect to cellular component, biological process, or molecular func-tion) is insufficient to capture differences between the groups. GOsummaries, for example, do not take into account relative quantita-tive amounts. Thus, different bioinformatic platforms and moredetailed analysis (e.g. pathway and network analysis) might detectcritical proteomic differences between the control and Pde6ɑD670G

retina at different stages. The GO distributions of the vitreous pro-teomes were distinct, illustrating that the vitreous and retina con-tained different functional categories of proteins with respect to theircondition. For example, arRP vitreous contained a higher proportionof membrane-bound proteins compared to wild-type (Fig. S2). Inaddition, the arRP retina contained a higher proportion of the GO cat-egory “antioxidant activity” compared to wild-type. Together, thisindicated that the arRP retina and vitreous express distinct functionalcategories of proteins compared to wild-type mice at each stage ofdisease progression, and that these protein pathways may reflectearly diagnostic markers or targets for therapeutics preventing fur-ther neuronal cell death and loss of the neural retinal network.

3.3. Oxidative metabolic pathways are downregulated in thePde6ɑD670G retina at the onset of neuronal cell death

We hypothesized that cellular pathways disrupted at the onset ofrod dystrophy might reflect key targets for therapeutic approaches.Therefore, we focused our further investigation on the Pde6ɑD670G

mouse proteomes collected at early-stage RP disease (P15). To examinedifferential protein expression at the onset of neuronal cell death, vitre-ous and retina protein levels (with two or more spectra) fromPde6ɑD670G (P15) and wild-type mice were compared using 1-wayANOVA and hierarchical heatmap clustering. A total of 1,067 proteinswere differentially expressed in the Pde6ɑD670G samples compared tocontrols (p < 0.05 [1-way ANOVA]; Fig. 4(a); Table S10). Similar to the

previous analysis comparing all Pde6ɑD670G stages (Fig. 3(c)), there was asignificant decrease in protein expression in Pde6ɑD670G retina at early-stage neurodegeneration (P15). There was also a significant increase invitreous protein expression at the onset of neuronal cell death in thePde6ɑD670G mice (779 upregulated proteins; p < 0.05 [1-way ANOVA]).The decrease in retinal protein expression and corresponding increasein vitreous protein expression suggested that the degenerating neuralretina may leak proteins into the vitreous, similar to that seen in humanpatient samples [23,41]. There were 446 proteins in the control retinathat were absent in the degenerating (Pde6ɑD670G) retina at P15. These446 proteins were absent in the control vitreous but, surprisingly, weredetected in the diseased (Pde6ɑD670G) vitreous. These results suggestthat 446 retinal proteins were depleted from the healthy retina andreleased into the vitreous at the onset of RP (Fig. 4(a); Table S11). In con-trast to the previous analysis comparing all Pde6ɑD670G stages (Fig. 3(c)),we observed a greater number of migrating proteins in the early stageanalysis (446 in Fig. 4(a) vs. 360 in Fig. 3(c)). This difference in observedproteins is due to the fact that fewer experimental groups are beingcompared in Fig. 4(a) (early stage) compared to Fig. 3(c) (all groups). GOanalysis categorized a significant fraction of these proteins to be intra-cellular, further suggesting that they have migrated from the degenerat-ing neural retina into the extracellular vitreous (Fig. S3a). The proteins inthis list with the most significant fold-change (FC) were ADP/ATP trans-locase 1 (SLC25A4; p-value = 1.68e�15 [1-way ANOVA]), malate dehy-drogenase 2 (MDH2; p-value = 1.00e�19 [1-way ANOVA]), and ATPsynthase (ATP5A1; p-value = 2.06e�13 [1-way ANOVA]). Categorizationof these 446 proteins by their biological process revealed that a signifi-cant fraction was involved in energetic processes (Fig. S3b).

We next analyzed the pathway representation of these vitreousbiomarkers. Pathway analysis revealed that these proteins repre-sent metabolic (e.g. OXPHOS, TCA cycle), synaptic signaling, andvisual transduction pathways (Fig. 4(b); Tables S12�15). As antici-pated, we found key proteins involved in the rod photoreceptortransduction pathway to be increased in the arRP vitreous at theonset of rod cell degeneration (P15): rhodopsin (RHO), guaninenucleotide-binding protein alpha and beta subunits (GNAT1 andGNB1), phosducin (PDC), cyclic nucleotide-gated channel beta 1(CNGB1), rhodopsin kinase (GRK1), recoverin (RCVRN), andS-arrestin (SAG). Taken together, these results suggest that these446 proteins are candidate biomarkers for early arRP detection andpossible targets for neuroprotective therapeutics. We identified pro-teins involved in glycolysis, TCA cycle, and OXPHOS: hexokinase(HK2), aldolase A, (ALDOA), phosphofructokinase (PFK1), enolase(ENO1), pyruvate kinase (PKM), pyruvate dehydrogenase E1 compo-nent subunit (PDHA1), citrate synthase (CS), aconitate hydratase

Fig. 3. Proteomic analysis of Pde6ɑD670G retina and vitreous at early-, mid-, and late-stage retinitis pigmentosa (RP). (a) Timeline of the clinical disease presentation in Pde6ɑD670G

mice. (b) Histological analysis of Pde6ɑD670G mouse retinas at early-(P11), mid-(P30) and late-(P72) stage disease compared to a wild-type mouse at (P12) shows loss of the ONL pho-toreceptor cells over time. RPE, retinal pigment epithelium; ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer. Scale bar = 100mm. (c) Protein spectral countswere analyzed by 1-way ANOVA followed by hierarchical clustering. A total of 1,384 proteins were differentially expressed in the Pde6ɑD670G vitreous and retina samples comparedto controls (p < 0.05 [1-way ANOVA]). Results are represented as a heatmap and display protein expression levels on a logarithmic scale. Orange indicates high expression whiledark green/black indicates low or no expression.


(ACO2), isocitrate dehydrogenase (IDH1 and �3), ɑ-ketoglutaratedehydrogenase (OGDH), succinate-CoA ligase beta subunit(SUCLA2), succinate dehydrogenase flavoprotein subunit (SDHA),and malate dehydrogenase (MDH1 and �2). We also identified pro-teins involved in fatty acid and amino acid synthesis pathways: fattyacid synthase (FASN) and glutamic-oxaloacetic transaminase (GOT1and GOT2). This observation suggested that retinal oxidative

metabolism was affected in early retinal degeneration. Defects inoxidative metabolism have been previously shown to contribute tophotoreceptor degeneration in RP patients [42]. For example, loss-of-function mutations in the IDH3B gene encoding the beta subunitof NAD-specific isocitrate dehydrogenase cause non-syndromic RP[43]. We similarly noted increased levels of FASN in the vitreous ofPde6ɑD670G mice at the onset of photoreceptor degeneration. Loss of

Fig. 4. Proteomic analysis of Pde6ɑD670G retina and vitreous identifies global protein expression changes. (a) Protein spectral counts were analyzed by 1-way ANOVA followed byhierarchical clustering. A total of 1,067 proteins were differentially expressed in the Pde6ɑD670G vitreous and retina samples (when comparing protein expression profiles at P15only) compared to controls (p < 0.05 [1-way ANOVA]). Results are represented as a heatmap and display protein expression levels on a logarithmic scale. Orange indicates highexpression while dark green/black indicates low or no expression. Pde6ɑD670G vitreous samples are highlighted by the red font. There were 446 proteins that were depleted in thePde6ɑD670G retina that were upregulated in the Pde6ɑD670G vitreous. There were 41 proteins that were depleted in the Pde6ɑD670G vitreous compared to the control vitreous. (b) Path-way representation of the 446 candidate biomarkers in the Pde6ɑD670G vitreous. Pathways are ranked by their fold-enrichment obtained from the Mann�Whitney U Test. Metabolicand energetic pathways are denoted by bold-faced text.


FASN in the neural retina has been shown to result in progressiveneurodegeneration resembling RP in mouse models [20]. Addition-ally, we have previously shown that metabolic and antioxidantstress proteins were regionally expressed within the human retina,in particular the OXPHOS and TCA cycle pathways, indicating thatthe retinal neural network is likely sensitive to changes within themetabolic pathways that occur during disease [11]. Although ourhuman case represented a late-stage of disease, after loss of the rodneuronal cells, the validation and identification of a candidate bio-marker with our early-stage Pde6ɑD670G mouse proteome suggestedpotential target pathways for therapeutic approaches (Fig. 5).

3.4. The ketogenic diet delays photoreceptor cell loss in arRP mice

We tested the arRP candidate biomarker, FASN, by targeting thiscritical protein with a dietary approach. The ketogenic diet hasgained public popularity for health and weight loss benefits, as wellas being an effective treatment option for young patients with epilep-tic seizures that do not respond to standard drug treatments [44]. Theketogenic diet provides acetyl-CoA that can feed into both the TCAcycle and the fatty acid synthesis pathways (Fig. 5). We placed arRPmice on a ketogenic diet (6:1:1 fat:protein:carbohydrate) beginningat postnatal day 21 (P21), mid-stage of photoreceptor degeneration.Histological analysis of the outer nuclear layer (ONL) thicknessshowed mild, but significant, photoreceptor cell survival in micetreated with the ketogenic diet compared to untreated control

littermates at one month of age (Fig. 6(a)). We next examined visualfunction at one month of age. Representative ERG traces of arRP micetreated with the ketogenic diet showed no difference in the rod pho-toreceptor cell response, but there was a mild rescue of the globalvisual response compared to littermates on standard chow diet(Fig. 6(b)). Quantification of ERG recordings from arRP mice treatedwith the ketogenic diet showed no significant difference in ERGresponses compared to littermates on a standard chow diet, althoughthere was a trend toward a- and b-wave visual rescue at the maximalscotopic 1.0 ERG setting (Fig. 6(c)�(e)). One limitation of the keto-genic diet was that the diet could not support nursing pups duringearly development, so it was not initiated until P21, at mid-stage ofretinal disease when a rescue effect may be diminished. Therefore,we considered oral supplementation of metabolites that could be ini-tiated earlier.

3.5. Oral supplementation with single metabolites prolongs neuronalcell survival and vision

The onset of retinal cell degeneration occurs rapidly in the arRPpreclinical model, before weaning age; therefore, metabolites werechosen that were known to transfer into the mother’s milk and crosstissue barriers to be fed to the nursing pups [45,46]. Thus, Pde6ɑD670G

mice were provided with ɑ-ketoglutarate (ɑ-KG), vitamin B3, vitaminB2, melatonin, or resveratrol in the drinking water beginning at post-natal day 0 (P0) and examined for functional visual rescue by ERG at

Fig. 5. Proteomics-guided metabolite supplementation strategy for arRP. Metabolic pathway diagram highlighting identified significantly elevated metabolic proteins in thePde6ɑD670G vitreous (red text). Metabolites are represented by blue text. Affected pathways include the TCA cycle, glycolysis, and oxidative phosphorylation (OXPHOS; magentatext). The metabolite/dietary therapies are highlighted in yellow.


one-month post-treatment (Fig. 5). Representative ERG tracesshowed visual rescue both in the rod photoreceptors, rod-cone pho-toreceptor a-wave response, and global inner retina scotopic ERGresponse in Pde6ɑD670G mice treated with ɑ-KG (Fig. 7). For vitaminB3 and vitamin B2, some mice showed enhanced ERG visual responsescompared to controls, but analysis of a larger cohort of mice did notshow statistical significance for photoreceptor cell and inner retinafunction on ERG (Fig. 7(b)�(d)). Both melatonin and resveratrol didnot show an effect on visual function after supplementation in thePde6ɑD670G arRP mice (Fig. 7(b)�(d)). Since ɑ-KG oral supplementa-tion prolonged visual function in the Pde6ɑD670G mice, we examinedthe retina by histology (Fig. 7(e)�(f)). Histological analysis confirmeda statistically significant rescue of photoreceptor cells (ONL) and theirinner/outer segments after treatment with ɑ-KG (Fig. 7(e)�(f)). It haspreviously been shown that glial fibrillary acidic protein (GFAP)expression is increased in M€uller glial processes that extend through-out the retina in response to retinal stress [47]. The distribution ofGFAP immunoreactivity was assessed to evaluate retinal response toinjury in untreated arRP mice and arRP mice treated with ɑ-KG. Bothgroups showed consistently increased GFAP immunoreactivity span-ning the length of the retina, suggesting that ɑ-KG supplementationdoes not inhibit M€uller glial activation in this disease model after onemonth of treatment (Fig. 7(g)). Overall, oral supplementation with asingle metabolite, ɑ-KG, provided a significant neuroprotective effecton the rod photoreceptors and neural retinal network for at least onemonth of age, when treated before disease onset. However, we can-not formally exclude the possibility that the systemic effects fromoral supplementation led to a neuroprotective outcome. The rescue

effects on photoreceptor cell survival and visual function after treat-ment with the ketogenic diet or ɑ-KG supported our human andmouse proteomic data, suggesting that the neural retina is sensitiveto changes in metabolism, and that targeting the TCA cycle may be apotential therapeutic approach for RP patients.

3.6. Metabolite/lipid imaging of retinal tissue sections

To investigate the lipid and metabolite profiles of Pde6ɑD670G micewith and without ɑ-KG treatment, we selected 35 fresh frozen tissuesamples and performed negative ion mode desorption electrosprayionization mass spectrometry imaging (DESI-MSI; Fig. 8(a)). Retinatissue samples were harvested from wild-type (n = 8), untreatedPde6ɑD670G (n = 8), and ɑ-KG treated Pde6ɑD670G mice (n = 14) at one-month of age. Additionally, whole eye (n = 5 for each group) sampleswere also harvested. For each case, 10 mm frozen sections were usedfor DESI-MSI. Each pixel (200mm) in the tissue DESI-MS image is rep-resented with the mass spectrum in the 50�1,000 m/z range. Themolecular ions identified were small metabolites related to glycoly-sis, TCA cycle, fatty acids, and phospholipids (Fig. 8(b)�(d)). The rela-tive intensity distribution of each individual metabolite in the wildtype, arRP, and treated tissue of both retina and the whole eye wereshown in 2D chemical heatmaps (Fig. S4-S5). Across 35 total tissuesamples, 26,341 unique peaks were found in the 50�1,000m/z range.Significance analysis of microarrays (SAM) identified 377 differen-tially expressed metabolites between untreated Pde6ɑD670G and ɑ-KGtreated mice. Although SAM returned 377 metabolites with notablefold-change (absolute fold change >1), to help interpret the

Fig. 6. Treatment with the ketogenic diet rescues photoreceptor cell survival in the arRP preclinical mouse. (a) Histological analysis of the retinas of Pde6ɑD670G mice on normalchow diet (gray) compared to Pde6ɑD670G mice provided the ketogenic diet (green) shows a significant increase in the thickness of the outer nuclear layer (ONL) by number of photo-receptor cell nuclei in mice treated with the ketogenic diet at one month of age, one-week post-dietary treatment. Results are displayed as a morphometric quantification (spidergraph) of ONL layer thickness superior (left) and inferior (right) to the optic nerve head (ONH). Wild-type mice have an ONL thickness of approximately 10 nuclei. ***p < 0.001 [Stu-dent’s t-test]. N = 8 eyes each group, with multiple ONL thickness counts per eye as described in Methods. Error bars = SEM. (b) Representative traces of the scotopic 0.1 (flashstrength in cd¢s¢m�2) dim-light rod-specific electroretinography (ERG; left) shows no significant difference in the rod photoreceptor cell response in Pde6ɑD670G mice treated withthe ketogenic diet (green) in comparison to untreated Pde6ɑD670G mice (gray) at one month of age. Representative traces of the scotopic 1.0 (flash strength in cd¢s¢m�2) global elec-troretinography (ERG; left) shows some rescue of the global visual response in Pde6ɑD670G mice treated with the ketogenic diet (green) in comparison to wild-type controls (black)and untreated Pde6ɑD670G mice (gray) at one month of age. Quantification of ERGs from a cohort of mice shows no significant visual rescue of (c) the rod photoreceptor cell response,(d) the maximum scotopic a-wave response, or (e) the maximum scotopic b-wave response in mice treated with the ketogenic diet (green) in comparison with wild-type controls(black) or untreated Pde6ɑD670G mice (gray) at one month of age, although there was a trend toward an increased a- and b-wave maximal response. N = 5 WT, 6 arRP, 6arRP + ketogenic diet mice.


biological role of metabolites, we restricted our analysis to peaks forwhich tandem-MS and high mass resolution analyses were per-formed by using the LTQ-Orbitrap XL (Thermo Scientific; Figs. S6�S7;Table S16). To check whether orally supplemented metabolites hadreached the retina, we first analyzed the changes in relative intensityfor ɑ-KG (m/z 145.014) in our DESI-MSI data. Although DESI-MSI diddetect ɑ-KG (Fig. S8) in some pixels, there was limited changebetween Pde6ɑD670G and ɑ-KG treated groups. This is possible if theorally supplemented ɑ-KG is being consumed for the production ofother metabolites in TCA and in glutamate-glutamine conversionpathways, which were ascertained next.

Identification of differentially distributed metabolites from SAMrevealed several altered pathways between untreated Pde6ɑD670G andɑ-KG treated groups that were consistent with our proteomic findings:uridine, dihydrouridine, and thymidine (pyrimidine and purinemetab-olism), docosahexaenoic and arachidonic acid (lipid metabolism), glu-tamine and glutamyl-alanine (glutamine/glutamate conversion andaminobutryate degradation), and ascorbic acid (oxidative metabolism).

We found intermediates in the TCA cycle that were highly elevated inɑ-KG treated mice compared to the untreated Pde6ɑD670G mice. Forexample, a-KG is broken down and reduced to succinyl-CoA, which isthen converted to succinic acid. From DESI-MSI imaging and SAM, wenoted that succinic acid (m/z 117.01) levels were elevated in thetreated state (fold change 1.3, FDR < 0.05) compared to the untreatedgroup (Fig. 9(a)). This indicated that the consumption of the supple-mental a-KG in the treatment group is reflected in higher levels of suc-cinic acid. Additionally, aconitic acid (m/z 173.0) is an intermediate inthe conversion of citrate to isocitrate in the TCA cycle, which can alsobe converted from ɑ-KG. Aconitic acid levels were elevated in thetreated state (fold change 2.5, FDR < 0.05), compared to wild-type anduntreated Pde6ɑD670G levels (Fig. 9(b)). Since it is known thatɑ-KG can act as a substrate for amino acid synthesis, particularly gluta-mine, we investigated whether the consumption of ɑ-KG was reflectedin glutamine synthesis pathways. Glutamine levels were elevated (foldchange 1.87, FDR < 0.05) in treated tissues compared to untreatedPde6ɑD670G mice (Fig. 9(c)). Glutamate levels were slightly increased in

Fig. 7. Oral supplementation of single metabolites rescues photoreceptor cell survival and visual function in the arRP preclinical mouse. (a) Representative traces of the scotopic 0.1(flash strength in cd¢s¢m�2) dim-light rod-specific electroretinography (ERG; left) and scotopic 1.0 (flash strength in cd¢s¢m�2) global ERG (right) in Pde6ɑD670G mice treated with ɑ-KG (blue), vitamin B2 (purple), vitamin B3 (red), melatonin (orange) or resveratrol (yellow) supplementation in comparison to untreated Pde6ɑD670G mice (gray) and wild-type con-trols (black) at one month of age. Quantification of ERGs from a cohort of mice treated with ɑ-KG (blue), vitamin B2 (purple), vitamin B3 (red), melatonin (orange) or resveratrol (yel-low) supplementation shows significant visual rescue of (b) the rod photoreceptor cell response, (c) the maximum scotopic a-wave response, and (d) the maximum scotopic b-waveresponse in mice treated with ɑ-KG supplementation (blue) in comparison with wild-type controls (black) or untreated Pde6ɑD670G mice (gray) at one month of age. **p < 0.01;***p < 0.001 [1-way ANOVA followed by Tukey’s multiple comparison’s test]. N = 5 WT, 6 arRP, 11 arRP + ɑ-KG mice, 8 arRP + vitamin B3, 7 arRP + vitamin B2, 8 arRP + melatoninand 7 arRP + resveratrol. (e-f) Histological analysis of the retinas of untreated Pde6ɑD670G mice (left) compared to treated Pde6ɑD670G mice (right) shows a significant increase in thethickness of the outer nuclear layer (ONL) by number of photoreceptor cell nuclei in mice treated with ɑ-KG supplementation (blue) at one month of age. GCL, ganglion cell layer;INL, inner nuclear layer; RPE, retinal pigment epithelium. Scale bar = 100 mm. (f) Results are displayed as a morphometric quantification (spider graph) of ONL layer thickness supe-rior (left) and inferior (right) to the optic nerve head (ONH). Wild-type mice have an ONL thickness of approximately 10 nuclei. ****p < 0.0001 [Student’s t-test]. N = 3 wild-type and4 arRP + ɑ-KG eyes, with multiple ONL thickness counts per eye as described in Methods. Error bars = SEM. (g) Extensive M€uller glial cell activation seen in arRP mice with and with-out ɑ-KG treatment. Retinas were dissected at one month of age from untreated arRP mice and arRP mice treated with ɑ-KG beginning at postnatal day 0. Retinas were stained withglial fibrillary acidic protein (GFAP; purple) and DAPI (blue) to identify retinal nuclei. In both the untreated arRP (left) and ɑ-KG treated (right) arRP mice there is increased expres-sion of GFAP in M€uller glia spanning the retina, suggesting that the treatment has no effect on Muller glial activation. N = 3 per group. Images taken at 25£ magnification. Scalebar = 100 mm.


Fig. 8. DESI-MSI of retinal tissue sections: (a) schematic representation of the DESI-MSI in the range m/z 50�1,000. The impinging primary charged droplets desorb the surfacemetabolites into the secondary charged droplets, followed by the ion transfer to the mass spectrometer. The 2D-chemical map of docosahexaenoic acid (DHA; peak listed in redfont) is shown with pixel size 200 mm. Raw MS spectrum across the pixels in a row is shown for (b) wild type retina tissue (c) arRP retina tissue and (d) ɑ-KG treated retina tissue.Red labels denote peaks assigned to identified metabolites and lipids. NL, normalization level; FA, fatty acid; PA, phosphatidic acid; PE, phosphatidylethanolamine; PG, phosphati-dylglycerol; PI, phosphatidylinositol; PS, phosphatidylserine.


treated tissues compared to untreated Pde6ɑD670G mice (Fig. 9(d)).Therefore, DESI-MSI provided some indication that supplemental ɑ-KGwas being consumed to increase levels of TCA metabolites such as aco-nitic acid and glutamine levels, which likely play a role in the cell’santioxidant response [48].

Consistent with the pathway representation of our vitreous prote-omic biomarkers, DESI-MSI independently detected altered levels ofmetabolites related to pyrimidine metabolism. Notably, our resultsfrom SAM revealed that levels of dihydrouridine (m/z 244.98) werelower in Pde6ɑD670G mice and restored following ɑ-KG treatment(fold change 18.1, FDR< 0.05; Fig. 9(e)). Similarly, levels of thymidine(m/z 241.17) and uridine (m/z 243.16) were elevated in ɑ-KG treatedmice compared to Pde6ɑD670G and wild-type mice, further suggestingthat ɑ-KG treatment influenced pyrimidine and purine metabolism(Fig. 9(f)�(g)). This trend of altered nucleotide metabolism in arRPhas been found in other murine metabolomics analyses [49]. Weisset al. had previously used broad-spectrum metabolomics to reportlinks between altered levels of thymidine and other pyrimidinenucleotide metabolic products and RP in a different RP murine model[49]. Overall, our work, in conjunction with other previously pub-lished efforts, suggests that pyrimidine metabolism is altered in RPand may be related to mitochondrial DNA replication pathways [48].

The proteomics data indicated the downregulation of oxidativepathways in Pde6ɑD670G retina compared to wild-type retina. Wetherefore looked to determine if any of the identified metabolites fromSAM reflected a change in oxidative pathways between the diseaseand treated groups. Ascorbic acid (m/z 175.06) was found at signifi-cantly higher intensities in untreated disease samples compared to thetreated group (fold change 1.9, FDR < 0.05; Fig. 9(h)). It is well-knownthat RP is characterized by rod photoreceptor cell death followed by

slower cone death, where cone death is associated with oxidativedamage [50]. Our DESI-MSI results revealed that the untreated diseasesamples had a higher distribution of ascorbic acid, which is a knownantioxidant. Additionally, there were metabolites that did not changesignificantly in the disease state but were elevated by ɑ-KG treatment.For example, docosahexaeonic acid (DHA;m/z 327.12) levels were ele-vated (fold change 1.9, FDR < 0.05) in the treated mice compared tountreated mice (Fig. 9(i)). DHA accounts for a significant fraction ofomega-3 fatty acids in the retina, and studies have found linksbetween reduction in DHA with altered retinal function, including inRP [50,51]. This result is consistent with the numerous suggested pro-tective roles of DHA in the retina, including anti-inflammatory, anti-angiogenesis, and anti-neurotoxic mechanisms [52].

4. Discussion

Liquid biopsies and the analysis of fluid compartments near diseasedtissues represents an innovative tool in precision health to overcome thecurrent limitations of tissue biopsies. This advancement is exemplified inthe personalized treatment of cancer, where circulating tumor-derivedmaterial (e.g. circulating tumor DNA/RNA, extracellular vesicles, andcells) are detected in the blood [53]. These cellular biomarkers of thetumor microenvironment aid clinicians to better determine the appro-priate therapeutic approach and predict prognosis. Neurodegenerativediseases, however, currently lack early detection methods for diagnosisand screening. In the case of retinal degenerative diseases, we have pre-viously shown that liquid biopsies of the adjacent vitreous fluid can cap-ture intracellular events in the neural retinal network that can guidediagnosis and treatment [23,41]. Advancements in this approach couldallow for serial sampling and longitudinal monitoring of disease progress

Fig. 9. Distributions of retinal lipids and metabolite relative intensities across wild-type, arRP, and ɑ-KG treated mice from DESI-MSI data: Distribution of (a) succinic acid (m/z117.01), (b) aconitic acid (m/z 173.0), (c) glutamine (m/z 145.0), (d) glutamate (m/z 146.0), (e) dihydrouridine (m/z 244.9), (f) uridine (m/z 243.1), (g) thymidine (m/z 241.1), (h) ascor-bic acid, and (i) docosahexaenoic acid (DHA; m/z 327.1) across wild-type, arRP, and ɑ-KG treated mice. Results are displayed as scatter plots of the relative intensity of the corre-sponding metabolite normalized by the total ion current (sum of all intensity values from the detected ions). Boxes represent the mean § SEM.


without the need to biopsy the neurosensory retina [24,36]. Our proteo-mic analysis of end-stage human arRP vitreous highlighted molecularpathways that might be affected during the progression of neuronal celldeath and retinal remodeling. In order to gain further insight into theearlier disease stages, we utilized a preclinical murine model of arRP.

We examined the retina and vitreous proteomes from Pde6ɑD670G

mice at early, mid, and late stages of neurodegeneration. Since therewas a large number of differentially expressed proteins betweenPde6ɑD670G mice and wild-type controls, we chose to focus on theproteins that were highly represented in the pathway analysis (Fig. 4(b)). The identification of related proteins within known pathways (e.g. TCA cycle, pyrimidine metabolism, phototransduction, etc.)allowed for focused interpretation of the data. In the early diseasestage, we observed rod phototransduction proteins were depletedin the retina and elevated within the vitreous before significant loss

of the photoreceptor cell nuclei is detectable by histological analysis.This suggests that molecular biomarkers are likely to appear in thehuman vitreous much earlier than the clinical biomarkers currentlyused for diagnosis. In human patients, liquid biopsies could similarlybe used as an early diagnostic to confirm the onset of photoreceptordegeneration, even without diagnosis of a specific genetic mutationby genetic testing. Additionally, we found that the vitreous pro-teomes differed in the arRP mouse model at early-, mid-, and late-stage of disease. It may be possible to use this liquid biopsy approachto estimate the molecular stage of disease based on the proteomiccontent of the vitreous. Comparison of proteomes before and aftertherapeutic approaches can also be studied using liquid biopsy andmay have important implications in gene therapy studies. Further,our DESI-MSI data complemented our understanding of key alteredpathways detected by our proteomics and pathway analysis


approach. Specifically, DESI-MSI provided a precise approach toinvestigate how changes at the protein level are reflected in the reti-nal metabolome of arRP mice following ɑ-KG treatment. Using DESI-MSI and SAM, we identified several significantly altered metabolitesbetween disease and treated groups, across multiple pathways,including the TCA cycle, glutamine synthesis, pyrimidine metabolism,oxidative pathways, and fatty acid metabolism. It is likely that DESI-MSI in tandem with studying the vitreous proteome can help identifypotential biomarkers that help elucidate effects of metabolite therapy.

At the early timepoint, the Pde6ɑD670G mice lose proteins criticalfor oxidative phosphorylation and aerobic metabolism, likelydepressing protein turnover and leading to the cell death of the rodphotoreceptors. Although the number of patient biopsies was limitedand the disease stage was late, there were overlapping proteins inthe vitreous proteome of the Pde6ɑD670G mice and humans. The lossof proteins involved in critical metabolic pathways from the retinawas supported by our previous proteomics of the human retina thathighlighted the sensitivity of the neural retinal network to changeswithin the metabolic pathways [11]. Interestingly, previous researchstudies have shown that metabolic rewiring is likely to be both acause and a consequence of photoreceptor degeneration [54], andthat the disruption of normal energy metabolism plays a key role inphotoreceptor degeneration in the rd10 RP mouse model [49]. Wealso found that there were significant alterations in purine andpyrimidine nucleotide metabolism pathways at disease onset in ourarRP mouse model, similar to the results in the rd10 RP mouse study(Table S12). Taken together, the proteomic profiles suggest targetingthe affected metabolic pathways by replenishing them though die-tary supplementation.

The most effective therapy tested was ɑ-KG, a key component ofthe TCA cycle that acts on various downstream metabolic pathwayswithin the cell and determines the overall rate of the energetic pro-cess [49] as well as enhancing reductive carboxylation and support-ing redox homeostasis in photoreceptors [55,56]. In addition to itsmetabolic roles, ɑ-KG can be converted to glutamine, and subse-quently glutathione GSH, thereby facilitating the synthesis of potentantioxidants in the cell [48]. These antioxidant properties may havebeneficial effects in the context of RP therapy. Additional studieshave shown that ɑ-KG supplementation can extend the lifespan of C.elegans through the inhibition of ATP synthase and TOR [57] as wellas increase circulating plasma levels of insulin and growth hormonesin humans [58]. In the mouse, ɑ-KG derivatives have been deliveredas a treatment for hypoxic conditions without any notations ofadverse side effects [59,60]. We found that oral supplementationwith ɑ-KG alone provided significant visual rescue of the rods, cones,and inner retina visual responses through at least one month of age.We chose the dosage of ɑ-KG based on published literature for sup-plementation in the mouse drinking water [61]. Additional studiesare needed to determine the lowest dosage of ɑ-KG and its durability.Previous studies in mice have demonstrated that high systemic dosesof ɑ-KG affected body weight and intestinal innate immunity byinfluencing the intestinal microbiome [61]. In this study, mice treatedwith ɑ-KG appeared to display a smaller body stature than controls,and higher doses of ɑ-KG would likely be detrimental to the survivalof the litters after birth. It is also possible that the rescue effect of ɑ-KG is indirect and a result of changes to the systemic metabolism.One way to distinguish between these effects in future studies wouldbe through intraocular administration of ɑ-KG.

The ketogenic diet has been shown to have neuroprotective effectsin the context of several neurodegenerative diseases, such as Alz-heimer’s and Parkinson’s disease [62]. The neuroprotective effect of theketogenic diet has been hypothesized to be a result of the biochemicalchanges resulting from glycolytic inhibition, increased acetyl-CoA pro-duction, and formation of ketone bodies in the cell [63]. These biochemi-cal changes have proven effective in targeting cancer cells, whichprimarily use glycolysis as their main source of ATP [64]. Photoreceptors

are similarly sensitive to glycolytic inhibition but can attenuate theimpact of glycolytic inhibition by using alternative sources of ATP (e.g.fatty acid beta-oxidation) [15,65]. The high rate of aerobic ATP synthesisin photoreceptors can lead to the production of harmful reactive oxygenspecies (ROS), which are implicated in the progression of retinal degen-erative diseases [42]. We have previously shown that different anatomi-cal regions of the human retina are susceptible to oxidative stress,highlighting the importance of antioxidant availability in the retina [11].Ketone bodies increase mitochondrial respiration via increased ATP pro-duction and reduce ROS formation by increasing NADH oxidation,thereby improving the efficiency of the mitochondrial respiratory chaincomplex [62]. Targeting the TCA cycle and fatty acid synthase pathwayby acetyl-CoA production via the ketogenic diet led to increased globalretinal visual function compared to controls by one month of age. As theketogenic diet could not be provided until weaning age, the effect waslimited based on beginning the dietary treatment after approximatelyhalf of the neuronal photoreceptor cells have already been lost. Never-theless, significant photoreceptor cell survival was detectable by histo-logical analysis after treatment with the ketogenic diet. This suggeststhat earlier delivery of metabolites targeting the TCA cycle may lead toprolonged visual rescue in the Pde6ɑD670G mouse model and neuropro-tection of the photoreceptor cells and inner retinal network.

Other metabolites tested in our arRP preclinical mouse modelwere less effective. Nicotinamide adenine dinucleotide (NAD), avail-able from vitamin B3 supplementation, is a critical co-enzyme for aer-obic metabolism and OXPHOS. In particular, retinal levels of NADdecrease with age and may leave neuronal cells susceptible to insultsand disease [66]. Furthermore, vitamin B3 supplementation hasshown positive effects toward enhancing energy metabolism andproper functioning of the central nervous system [67] as well as act-ing as a protective agent against glaucoma and age-related neurode-generation [66]. Flavin mononucleotide (FMN) and flavin adeninedinucleotide (FAD), which can be delivered via vitamin B2 supple-mentation, are critical for enzymatic reactions within the electrontransport chain and the TCA cycle. Vitamin B2 acts as an antioxidantby promoting regeneration of glutathione and is indispensable forcellular growth and supplementation with vitamin B2 has shown pos-sible beneficial effects for prevention of hyperhomocysteinemia, cat-aracts and migraine headaches [68]. We found that targetingOXPHOS by oral supplementation with vitamins B2 or B3 provided anelectrophysiological rescue effect in some, but not all, neuronal cellsof treated mice compared to their respective controls. Further studiesare needed to test combination therapy with our effective treatmentgroups (ɑ-KG, the ketogenic diet, and/or vitamins B2 and B3) that mayhave a synergistic effect in rescuing the neuronal cell loss and moreclosely reproduce a strategy to use in human arRP patients. Longer-acting structural analogs of these metabolites may also provide a lon-ger lasting effect.

We also tested metabolites that act on the sirtuin pathway andmay indirectly affect cellular metabolism. Melatonin is producedwithin the retina and is known to stimulate sirtuin activity in mito-chondria, which deacetylates many of the enzymes involved withinthe TCA cycle. Previous studies have shown that supplementationwith melatonin has no known toxicity affects when taken duringpregnancy in both humans and mice [69,70]. Additionally, resveratrolis a plant-derived polyphenol that has been widely studied for itspotential benefits as a protective agent against cancers, mutagenesis,and a variety of other biological functions [70]. It also acts to stimu-late the sirtuin pathway, and this stimulation as well as its antioxida-tive actions are thought to provide a neuroprotective effect in bothrat and mouse studies [70]. Supplementation with resveratrol has noknown toxicity in humans even at high dosages, and the only affectnoted when taken during pregnancy in mice was a slight increase inthe body weight of the pups after birth [71]. However, in our study,oral supplementation with either melatonin or resveratrol did notprovide a neuroprotective effect on the photoreceptor cells in our


preclinical model of arRP. This suggests that directly targeting theTCA cycle may be the most beneficial approach for future RP therapy.

The proteomic and DESI-MSI approach in this study provided keypathways for targeting and testing therapeutics, and delivery ofmetabolites within these critical pathways in the arRP preclinicalmouse model showed efficacious neuroprotective effects on the innerretina. The results from these studies lay the groundwork for futureexperiments to address how the TCA cycle and aerobic metabolisminfluence the photoreceptors and their signaling to the inner retinalnetwork, and thus improve upon neuroprotective approaches target-ing photoreceptor cell metabolism. Overall, the key metabolic path-ways (OXPHOS and TCA cycle) detected in our proteomics screen arecritical targets for therapeutics for PDE6-related RP. Further investi-gation is required to determine if this approach could be applicableto other forms of RP or other human neurodegenerative disorders.

Funding sources

VBM and AGB are supported by NIH grants [R01EY026682,R01EY024665, R01EY025225, R01EY024698, R21AG050437 andP30EY026877], and Research to Prevent Blindness (RPB), New York, NY.GV is supported by NIH grants [F30EYE027986 and T32GM007337].Jonas Children’s Vision Care and Bernard & Shirlee Brown GlaucomaLaboratory are supported by the National Institute of Health[5P30EY019007, R01EY018213], National Cancer Institute Core[5P30CA013696], the Research to Prevent Blindness (RPB) Physician-Scientist Award, unrestricted funds from RPB, New York, NY, USA. Foun-dation Fighting Blindness [TA-NMT-0116�0692-COLU], the Research toPrevent Blindness (RPB) Physician-Scientist Award, and unrestrictedfunds from RPB, New York, NY, USA. S.H.T. is a member of the RD-CUREConsortium and is supported by Kobi and Nancy Karp, the CrowleyFamily Fund, the Rosenbaum Family Foundation, the Tistou and Char-lotte Kerstan Foundation, the Schneeweiss Stem Cell Fund, New YorkState [C029572], and the Gebroe Family Foundation. RNZ is supportedby NSF grant CHE-1734082.

Data and materials availability

All data associated with this study are present in the paper or theSupplementary Materials. The mass spectrometry proteomics datahave been deposited to Mendeley Data with the dataset identifiers(DOI): 10.17632/x9mkhv743y.2 and 10.17632/vn5c8zh5wd.2.

Declaration of Competing Interest

Dr. Wert has nothing to disclose. Mr. Velez has nothing to disclose.Dr. Kanchustambham has nothing to disclose. Mr. Shankar has noth-ing to disclose. Ms. Evans has nothing to disclose. Dr. Sengillo hasnothing to disclose. Dr. Zare has nothing to disclose. Dr. Bassuk hasnothing to disclose. Dr. Tsang has nothing to disclose. Dr. Mahajanhas nothing to disclose.

Acknowledgments

We thank Jing Yang, Daniel A. Machlab, and Teja Chemudupati fortechnical assistance. We thank Bioproximity for assistance with massspectrometry data collection. We thank Patsy Nishina for mousemodel resources. We thank Jianhai Du and his laboratory membersfrom the University of Washington for advice and discussions.

Supplementary materials

Supplementary material associated with this article can be foundin the online version at doi:10.1016/j.ebiom.2020.102636.

References

[1] Procaccini C, Santopaolo M, Faicchia D, Colamatteo A, Formisano L, de Candia P,et al. Role of metabolism in neurodegenerative disorders. Metabolism 2016;65(9):1376–90.

[2] Lehtonen S, Sonninen TM, Wojciechowski S, Goldsteins G, Koistinaho J. Dysfunc-tion of cellular proteostasis in Parkinson's disease. Front Neurosci 2019;13:457.

[3] Strohm L, Behrends C. Glia-specific autophagy dysfunction in als. Semin Cell DevBiol 2019.

[4] Olivares-Banuelos TN, Chi-Castaneda D, Ortega A. Glutamate transporters: geneexpression regulation and signaling properties. Neuropharmacology 2019.

[5] Shukla M, Chinchalongporn V, Govitrapong P, Reiter RJ. The role of melatonin intargeting cell signaling pathways in neurodegeneration. Ann NY Acad Sci2019;1443(1):75–96.

[6] Ambrogini P, Torquato P, Bartolini D, Albertini MC, Lattanzi D, Di Palma M, et al.Excitotoxicity, neuroinflammation and oxidant stress as molecular bases of epi-leptogenesis and epilepsy-derived neurodegeneration: the role of vitamin E. Bio-chim Biophys Acta Mol Basis Dis 2019;1865(6):1098–112.

[7] Chi H, Chang HY, Sang TK. Neuronal cell death mechanisms in major neurodegen-erative diseases. Int J Mol Sci 2018;19(10).

[8] Tsang SH, Sharma T. Stargardt disease. Adv Exp Med Biol 2018;1085:139–51.[9] Luo YH, da Cruz L. The Argus((R)) II retinal prosthesis system. Prog Retin Eye Res

2016;50:89–107.[10] Maguire AM, Simonelli F, Pierce EA, Pugh Jr. EN, Mingozzi F, Bennicelli J, et al.

Safety and efficacy of gene transfer for Leber's congenital amaurosis. N Engl J Med2008;358(21):2240–8.

[11] Velez G, Machlab DA, Tang PH, Sun Y, Tsang SH, Bassuk AG, et al. Proteomic analy-sis of the human retina reveals region-specific susceptibilities to metabolic- andoxidative stress-related diseases. PLoS ONE 2018;13(2):e0193250.

[12] Wang L, Tornquist P, Bill A. Glucose metabolism in pig outer retina in light anddarkness. Acta Physiol Scand 1997;160(1):75–81.

[13] Wang L, Tornquist P, Bill A. Glucose metabolism of the inner retina in pigs in dark-ness and light. Acta Physiol Scand 1997;160(1):71–4.

[14] Graymore C. Possible significance of the isoenzymes of lactic dehydrogenase inthe retina of the rat. Nature 1964;201:615–6.

[15] Noell WK. The effect of iodoacetate on the vertebrate retina. J Cell Comp Physiol1951;37(2):283–307.

[16] Chertov AO, Holzhausen L, Kuok IT, Couron D, Parker E, Linton JD, et al. Roles ofglucose in photoreceptor survival. J Biol Chem 2011;286(40):34700–11.

[17] Narayan DS, Chidlow G, Wood JP, Casson RJ. Glucose metabolism in mammalianphotoreceptor inner and outer segments. Clin Exp Ophthalmol 2017;45(7):730–41.

[18] Vaughn AE, DeshmukhM. Glucosemetabolism inhibits apoptosis in neurons and can-cer cells by redox inactivation of cytochrome C. Nat Cell Biol 2008;10(12):1477–83.

[19] Robertson JP, Faulkner A, Vernon RG. Regulation of glycolysis and fatty acid syn-thesis from glucose in sheep adipose tissue. Biochem J 1982;206(3):577–86.

[20] Rajagopal R, Zhang S, Wei X, Doggett T, Adak S, Enright J, et al. Retinal de novolipogenesis coordinates neurotrophic signaling to maintain vision. JCI Insight2018;3(1).

[21] Joyal JS, Gantner ML, Smith LEH. Retinal energy demands control vascular supplyof the retina in development and disease: the role of neuronal lipid and glucosemetabolism. Prog Retin Eye Res 2018;64:131–56.

[22] Zhang L, Du J, Justus S, Hsu CW, Bonet-Ponce L, Wu WH, et al. Reprogrammingmetabolism by targeting sirtuin 6 attenuates retinal degeneration. J Clin Invest2016;126(12):4659–73.

[23] Velez G, Roybal CN, Colgan D, Tsang SH, Bassuk AG, Mahajan VB. Precision medi-cine: personalized proteomics for the diagnosis and treatment of idiopathicinflammatory disease. JAMA Ophthalmol 2016;134(4):444–8.

[24] Mahajan VB, Skeie JM. Translational vitreous proteomics. Proteomics Clin Appl2014;8(3�4):204–8.

[25] Rowell HA, Bassuk AG, Mahajan VB. Monozygotic twins with CAPN5 autosomaldominant neovascular inflammatory vitreoretinopathy. Clin Ophthalmol2012;6:2037–44.

[26] Mahajan VB, Skeie JM, Bassuk AG, Fingert JH, Braun TA, Daggett HT, et al. Calpain-5 mutations cause autoimmune uveitis, retinal neovascularization, and photore-ceptor degeneration. PLoS Genet 2012;8(10):e1003001.

[27] Skeie JM, Brown EN, Martinez HD, Russell SR, Birkholz ES, Folk JC, et al. Proteomicanalysis of vitreous biopsy techniques. Retina 2012;32(10):2141–9.

[28] Wert KJ, Davis RJ, Sancho-Pelluz J, Nishina PM, Tsang SH. Gene therapy provideslong-term visual function in a pre-clinical model of retinitis pigmentosa. HumMol Genet 2013;22(3):558–67.

[29] Wert KJ, Sancho-Pelluz J, Tsang SH. Mid-stage intervention achieves similar effi-cacy as conventional early-stage treatment using gene therapy in a pre-clinicalmodel of retinitis pigmentosa. HumMol Genet 2014;23(2):514–23.

[30] Wert KJ, Bassuk AG, Wu WH, Gakhar L, Coglan D, Mahajan M, et al. CAPN5 muta-tion in hereditary uveitis: the R243L mutation increases calpain catalytic activityand triggers intraocular inflammation in a mouse model. Hum Mol Genet2015;24(16):4584–98.

[31] Wert KJ, Skeie JM, Bassuk AG, Olivier AK, Tsang SH, Mahajan VB. Functional vali-dation of a human CAPN5 exome variant by lentiviral transduction into mouseretina. HumMol Genet 2014;23(10):2665–77.

[32] Skeie JM, Tsang SH, Mahajan VB. Evisceration of mouse vitreous and retina forproteomic analyses. J Vis Exp 2011(50).

[33] Mi H, Muruganujan A, Ebert D, Huang X, Thomas PD. PANTHER version 14: moregenomes, a new panther GO-slim and improvements in enrichment analysistools. Nucleic Acids Res 2019;47(D1):D419–D26.


http://refhub.elsevier.com/S2352-3964(20)30011-6/sbref0001





















































































[34] Tusher VG, Tibshirani R, Chu G. Significance analysis of microarrays applied to theionizing radiation response. Proc Natl Acad Sci U S A 2001;98(9):5116–21.

[35] Tsang SH, Tsui I, Chou CL, Zernant J, Haamer E, Iranmanesh R, et al. A novel muta-tion and phenotypes in phosphodiesterase 6 deficiency. Am J Ophthalmol2008;146(5):780–8.

[36] Skeie JM, Roybal CN, Mahajan VB. Proteomic insight into the molecular functionof the vitreous. PLoS ONE 2015;10(5):e0127567.

[37] Azar WJ, Azar SH, Higgins S, Hu JF, Hoffman AR, Newgreen DF, et al. IGFBP-2enhances VEGF gene promoter activity and consequent promotion of angiogene-sis by neuroblastoma cells. Endocrinology 2011;152(9):3332–42.

[38] Park SH, Kim KW, Kim JC. The role of insulin-like growth factor binding protein 2(IGFBP2) in the regulation of corneal fibroblast differentiation. Invest OphthalmolVis Sci 2015;56(12):7293–302.

[39] Afshari FT, Kwok JC, Andrews MR, Blits B, Martin KR, Faissner A, et al. Integrinactivation or alpha 9 expression allows retinal pigmented epithelial cell adhesionon Bruch's membrane in wet age-related macular degeneration. Brain 2010;133(Pt 2):448–64.

[40] Malissen B, Schmitt-Verhulst AM. Transmembrane signalling through the T-cell-receptor-CD3 complex. Curr Opin Immunol 1993;5(3):324–33.

[41] Velez G, Tang PH, Cabral T, Cho GY, Machlab DA, Tsang SH, et al. Personalized pro-teomics for precision health: identifying biomarkers of vitreoretinal disease.Transl Vis Sci Technol 2018;7(5):12.

[42] Punzo C, Xiong W, Cepko CL. Loss of daylight vision in retinal degeneration: areoxidative stress and metabolic dysregulation to blame? J Biol Chem 2012;287(3):1642–8.

[43] Hartong DT, Dange M, McGee TL, Berson EL, Dryja TP, Colman RF. Insights fromretinitis pigmentosa into the roles of isocitrate dehydrogenases in the Krebs cycle.Nat Genet 2008;40(10):1230–4.

[44] D'Andrea Meira I, Romao TT, Pires do Prado HJ, Kruger LT, Pires MEP, da ConceicaoPO. Ketogenic diet and epilepsy: what we know so far. Front Neurosci 2019;13:5.

[45] Smilowitz JT, O'Sullivan A, Barile D, German JB, Lonnerdal B, Slupsky CM. Thehuman milk metabolome reveals diverse oligosaccharide profiles. J Nutr2013;143(11):1709–18.

[46] Wishart DS, Feunang YD, Marcu A, Guo AC, Liang K, Vazquez-Fresno R, et al.HMDB 4.0: the human metabolome database for 2018. Nucleic Acids Res 2018;46(D1):D608–D17.

[47] Lewis GP, Fisher SK. Up-regulation of glial fibrillary acidic protein in response toretinal injury: its potential role in glial remodeling and a comparison to vimentinexpression. Int Rev Cytol 2003;230:263–90.

[48] Liu S, He L, Yao K. The antioxidative function of alpha-ketoglutarate and its appli-cations. Biomed Res Int 2018;2018:3408467.

[49] Weiss ER, Osawa S, Xiong Y, Dhungana S, Carlson J, McRitchie S, et al. Broad spec-trum metabolomics for detection of abnormal metabolic pathways in a mousemodel for retinitis pigmentosa. Exp Eye Res 2019;184:135–45.

[50] Komeima K, Rogers BS, Lu L, Campochiaro PA. Antioxidants reduce cone cell deathin a model of retinitis pigmentosa. Proc Natl Acad Sci U S A 2006;103(30):11300–5.

[51] Hoffman DR, DeMar JC, Heird WC, Birch DG, Anderson RE. Impaired synthesisof DHA in patients with X-linked retinitis pigmentosa. J Lipid Res 2001;42(9):1395–401.

[52] Querques G, Forte R, Souied EH. Retina and omega-3. J Nutr Metab 2011;2011:748361.

[53] De Rubis G, Rajeev Krishnan S, Bebawy M. Liquid biopsies in cancer diagnosis,monitoring, and prognosis. Trends Pharmacol Sci 2019;40(3):172–86.

[54] Du J, An J, Linton JD, Wang Y, Hurley JB. How excessive cGMP impacts metabolicproteins in retinas at the onset of degeneration. Adv Exp Med Biol 2018;1074:289–95.

[55] Jiang L, Shestov AA, Swain P, Yang C, Parker SJ, Wang QA, et al. Reductive carbox-ylation supports redox homeostasis during anchorage-independent growth.Nature 2016;532(7598):255–8.

[56] Mullen AR, Wheaton WW, Jin ES, Chen PH, Sullivan LB, Cheng T, et al. Reductivecarboxylation supports growth in tumour cells with defective mitochondria.Nature 2011;481(7381):385–8.

[57] Chin RM, Fu X, Pai MY, Vergnes L, Hwang H, Deng G, et al. The metabolite alpha-ketoglutarate extends lifespan by inhibiting ATP synthase and TOR. Nature2014;510(7505):397–401.

[58] Wu N, Yang M, Gaur U, Xu H, Yao Y, Li D. Alpha-Ketoglutarate: physiological func-tions and applications. Biomol Ther (Seoul) 2016;24(1):1–8.

[59] Hou P, Kuo CY, Cheng CT, Liou JP, Ann DK, Chen Q. Intermediary metabolite pre-cursor dimethyl-2-ketoglutarate stabilizes hypoxia-inducible factor-1alpha byinhibiting prolyl-4-hydroxylase PHD2. PLoS ONE 2014;9(11):e113865.

[60] MacKenzie ED, Selak MA, Tennant DA, Payne LJ, Crosby S, Frederiksen CM, et al.Cell-permeating alpha-ketoglutarate derivatives alleviate pseudohypoxia in suc-cinate dehydrogenase-deficient cells. Mol Cell Biol 2007;27(9):3282–9.

[61] Chen S, Bin P, Ren W, Gao W, Liu G, Yin J, et al. Alpha-ketoglutarate (AKG) lowersbody weight and affects intestinal innate immunity through influencing intestinalmicrobiota. Oncotarget 2017;8(24):38184–92.

[62] Wlodarek D. Role of ketogenic diets in neurodegenerative diseases (Alzheimer’sdisease and Parkinson’s disease). Nutrients 2019;11(1).

[63] Puchalska P, Crawford PA. Multi-dimensional roles of ketone bodies in fuelmetabolism, signaling, and therapeutics. Cell Metab 2017;25(2):262–84.

[64] Weber DD, Aminazdeh-Gohari S, Kofler B. Ketogenic diet in cancer therapy. Aging(Albany NY) 2018;10(2):164–5.

[65] Joyal JS, Sun Y, Gantner ML, Shao Z, Evans LP, Saba N, et al. Retinal lipid and glu-cose metabolism dictates angiogenesis through the lipid sensor Ffar1. Nat Med2016;22(4):439–45.

[66] Williams PA, Harder JM, Foxworth NE, Cochran KE, Philip VM, Porciatti V, et al.Vitamin B3 modulates mitochondrial vulnerability and prevents glaucoma inaged mice. Science 2017;355(6326):756–60.

[67] Mills KF, Yoshida S, Stein LR, Grozio A, Kubota S, Sasaki Y, et al. Long-Term admin-istration of nicotinamide mononucleotide mitigates age-associated physiologicaldecline in mice. Cell Metab 2016;24(6):795–806.

[68] Peechakara BV, Gupta M. Vitamin B2 (Riboflavin). StatPearls. Treasure Island (FL)2019.

[69] Guan S, Xie L, Ma T, Lv D, Jing W, Tian X, et al. Effects of melatonin on early preg-nancy in mouse: involving the regulation of StAR, Cyp11a1, and Ihh expression.Int J Mol Sci 2017;18(8).

[70] Ramis MR, Esteban S, Miralles A, Tan DX, Reiter RJ. Caloric restriction, resveratroland melatonin: role of SIRT1 and implications for aging and related-diseases.Mech Ageing Dev 2015;146�148:28–41.

[71] Zheng S, Feng Q, Cheng J, Zheng J. Maternal resveratrol consumption and its pro-gramming effects on metabolic health in offspring mechanisms and potentialimplications. Biosci Rep 2018;38(2).






































































































metabolite therapy guided by liquid biopsy proteomics ... · h department of pathology & cell...

Documents